Integrated networking, caching, and computing

applica-of mobile cloud computing MCC [55], which leverages the rich tion resources in the cloud for user computation task execution, by enablingcomputation offloading through wireless c

Integrated Networking, Caching,

and Computing

Integrated Networking, Caching,

and Computing

F Richard Yu Tao Huang Yunjie Liu

© 2018 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S Government works

Printed on acid-free paper

Version Date: 20180428

International Standard Book Number-13: 978-1-138-08903-7 (Hardback)

This book contains information obtained from authentic and highly regarded sources Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint.

Except as permitted under U.S Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access

www.copyright.com ( http://www.copyright.com/ ) or contact the Copyright Clearance Center, Inc (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400 CCC is a not-for-profit organization that provides licenses and registration for a variety of users For organizations that have been granted

a photocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and

are used only for identification and explanation without intent to infringe.

Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com

and the CRC Press Web site at

http://www.crcpress.com

1 Overview, Motivations and Frameworks 1

1.1 Overview 2

1.1.1 Recent advances in wireless networking 2

1.1.2 Caching 4

1.1.3 Computing 5

1.2 Motivations and requirements 7

1.2.1 What is integration of networking, caching and computing? 7

1.2.2 Why do we need integration of networking, caching and computing? 8

1.2.2.1 The growth of networking alone is not sustain-able 8

1.2.2.2 The benefits brought by the integration of net-working, caching and computing 10

1.2.3 The requirements of integration of networking, caching and computing 11

1.2.3.1 Coexistence 11

1.2.3.2 Flexibility 11

1.2.3.3 Manageability and programmability 11

1.2.3.4 Heterogeneity 12

1.2.3.5 Scalability 12

1.2.3.6 Stability and convergence 12

1.2.3.7 Mobility 12

1.2.3.8 Backward compatibility 12

1.3 Frameworks 12

1.3.1 Caching-networking framework 13

1.3.1.1 D2D delivery (Fig 1.2a) 14

1.3.1.2 Multihop delivery via D2D relay (Fig 1.2b) 14

1.3.1.4 Direct SBS delivery (Fig 1.2d) 14

1.3.1.5 Cooperative SBS delivery (Fig 1.2e) 14

1.3.2 Computing-networking framework 18

1.3.2.1 Cloud mobile media 18

1.3.2.2 Mobile edge computing 19

1.3.3 Caching-computing framework 19

1.3.4 Caching-computing-networking framework 22

1.3.4.1 Networking-caching-computing convergence 22 1.3.4.2 Networking and computing assisted caching 23 1.3.5 A use case 23

References 25

2 Performance Metrics and Enabling Technologies 33

2.1 Performance metrics 33

2.1.1 General metrics 33

2.1.1.1 Cost 33

2.1.1.2 Revenue 36

2.1.1.3 Recovery time 36

2.1.2 Networking-related metrics 36

2.1.2.1 Coverage and capacity (throughput) 36

2.1.2.2 Deployment efficiency 36

2.1.2.3 Spectral efficiency 37

2.1.2.4 Energy efficiency 37

2.1.2.5 QoS 37

2.1.2.6 Signaling delay and service latency 37

2.1.3 Caching-related metrics 38

2.1.3.1 Average latency 38

2.1.3.2 Hop-count 38

2.1.3.3 Load fairness 38

2.1.3.4 Responses per request 39

2.1.3.5 Cache hits 39

2.1.3.6 Caching efficiency 39

2.1.3.7 Caching frequency 39

2.1.3.8 Cache diversity 39

2.1.3.9 Cache redundancy 39

2.1.3.10 Absorption time 39

2.1.4 Computing-related metrics 39

2.1.4.1 Execution time 39

2.1.4.2 Energy consumption 40

2.1.4.3 Computation dropping cost 40

2.1.4.4 Throughput 40

2.2 Enabling technologies 40

2.2.1 Caching-networking 41

2.2.1.1 Caching in heterogeneous networks 41

2.2.1.2 Caching in information-centric networking 42 2.2.1.3 Caching in D2D networking 43

2.2.1.4 Others 44

2.2.2 Computing-networking 44

2.2.2.1 Cloud computing and networking 44

2.2.2.2 Fog computing and networking 46

2.2.2.3 Mobile edge computing and networking 47

2.2.3 Caching-computing-networking 49

References 58

3 Edge Caching with Wireless Software-Defined Networking 65 3.1 Wireless SDN and edge caching 66

3.1.1 Motivations and contributions 66

3.1.2 Literature review 67

3.2 System model and problem formulation 68

3.2.1 Network Model 68

3.2.1.1 Wireless communication model 68

3.2.1.2 Proactive wireless edge caching model 71

3.2.1.3 Video QoE model 72

3.2.2 Problem formulation 73

3.3 Bandwidth provisioning and edge caching 75

3.3.1 Proposed caching decoupling via dual decomposition 76 3.3.2 Upper bound approach to solving (3.14) 77

3.3.3 Rounding methods based on marginal benefits 79

3.3.4 Computational complexity, convergence and optimality 80 3.3.5 Implementation design in SDWNs 82

3.4 Simulation results and discussion 83

3.4.1 Algorithm performance 84

3.4.2 Network performance 86

3.4.2.1 Delay 86

3.4.2.2 QoE guarantee 88

3.4.3 Utilization 88

3.4.3.1 Caching resources 88

3.4.3.2 Backhaul resource 89

3.5 Conclusions and future work 90

References 90

4 Resource Allocation for 3C-Enabled HetNets 95

4.1 Introduction 96

4.2 Architecture overview 98

4.2.1 Wireless network virtualization 98

4.2.2 Information-centric networking 98

4.2.3 Mobile edge computing 99

4.3 Virtualized multi-resources allocation 102

4.3.1 System model 102

4.3.1.1 Virtual heterogeneous networks model 102

4.3.1.2 Computing model 102

4.3.1.3 Caching model 105

4.3.2 Problem formulation 106

4.3.3 Problem reformulation 107

4.4 Resource allocation via ADMM 109

4.4.1 Decoupling of association indicators 109

4.4.2 Problem solving via ADMM 110

4.4.3 Algorithm analysis: computational complexity 113

4.5 Simulation results and discussion 113

4.5.1 Parameter settings 114

4.5.2 Alternative schemes 115

4.5.3 Performance evaluation 115

4.6 Conclusion and future work 121

References 122

5 Network Slicing and Caching in 5G Cellular Networks 125

5.1 Introduction 126

5.2 System model and problem formulation 128

5.2.1 Overview of a 5G core network involving network slicing and caching 129

5.2.2 System model and problem formulation 130

5.3 Caching resource allocation based on the CRO algorithm 133

5.3.1 Brief introduction to the CRO algorithm 134

5.3.2 Caching resource allocation based on the CRO algorithm 134

5.3.3 Complexity analysis 138

5.4 Simulation results and discussions 139

5.5 Conclusions and future work 144

References 144

6 Joint optimization of 3C 149

6.1 Introduction 149

6.2 System model 151

6.2.1 Network model 151

6.2.2 Communication model 154

6.2.3 Computation model 154

6.2.3.1 Local computing 155

6.2.3.2 MEC server computing 155

6.2.4 Caching model 156

6.2.5 Utility function 156

6.3 Problem formulation, transformation and decomposition 158

6.3.1 Problem formulation 158

6.3.2 Problem transformation 159

6.3.2.1 Binary variable relaxation 160

6.3.2.2 Substitution of the product term 160

6.3.3 Convexity 161

6.3.4 Problem decomposition 162

6.4 Problem solving via ADMM 164

6.4.1 Augmented Lagrangian and ADMM sequential itera-tions 164

6.4.2 Local variables update 166

6.4.3 Global variables and Lagrange multipliers update 167

6.4.4 Algorithm stopping criterion and convergence 169

6.4.5 Binary variables recovery 169

6.4.6 Feasibility, complexity and summary of the algorithm 170 6.5 Simulation results and discussion 172

6.6 Conclusions and future work 179

References 179

7 Software-Defined Networking, Caching and Computing 185

7.1 Introduction 186

7.2 Recent advances in networking, caching and computing 188

7.2.1 Software-defined networking 188

7.2.2 Information centric networking 189

7.2.3 Cloud and fog computing 189

7.2.4 An integrated framework for software-defined network-ing, caching and computing 190

7.2.4.1 Software-defined and information-centric control 190

7.2.4.2 Service-oriented request/reply paradigm 190

7.2.4.3 In-network caching and computing 191

7.3 Architecture of the integrated framework SD-NCC 191

7.3.1 The data plane 191

7.3.2 The control plane 193

7.3.3 The management plane 197

7.3.4 The workflow of SD-NCC 198

7.4 System model 200

7.4.1 Network model 200

7.4.2 Caching/computing model 200

7.4.3 Server selection model 201

7.4.4 Routing model 201

7.4.5 Energy model 201

7.4.5.1 Caching energy 201

7.4.5.2 Computing energy 202

7.5 Caching/computing/bandwidth resource allocation 202

7.5.1 Problem formulation 203

7.5.1.1 Objective function 203

7.5.1.2 Formulation 203

7.5.2 Caching/computing capacity allocation 204

7.5.3 The exhaustive-search algorithm 205

7.6 Simulation results and discussion 206

7.6.1 Network usage cost 206

7.6.2 Energy consumption cost 207

7.6.3 Optimal deployment numbers 208

7.7 Open research issues 209

7.7.1 Scalable SD-NCC controller design 209

7.7.2 Local autonomy in the SD-NCC data plane 210

7.7.3 Networking/caching/computing resource allocation strategies 210

7.8 Conclusions 211

References 211

8 Challenges and Broader Perspectives 215

8.1 Challenges 215

8.1.1 Stringent latency requirements 215

8.1.2 Tremendous amount of data against network bandwidth constraints 216

8.1.3 Uninterruptable services against intermittent connectivity 216

8.1.4 Interference of multiple interfaces 216

8.1.5 Network effectiveness in the face of mobility 217

8.1.6 The networking-caching-computing capacity 218

8.1.7 The networking-caching-computing tradeoffs 218

8.1.8 Security 218

8.1.9 Convergence and consistency 219

8.1.10 End-to-end architectural tradeoffs 219

8.2 Broader perspectives 219

8.2.1 Software-defined networking 219

8.2.2 Network function virtualization 220

8.2.3 Wireless network virtualization 221

8.2.4 Big data analytics 221

8.2.5 Deep reinforcement learning 223

References 223

Index 227

Overview, Motivations and Frameworks

New mobile applications, such as natural language processing, augmented ality and face recognition [1], have emerged rapidly in recent years However,conventional wireless networks solely focusing on communication are no longercapable of meeting the demand raised by such applications not only on highdata rates, but also on high caching and computing capabilities [2] Althoughthe pivotal role of communication in wireless networks can never be overem-phasized, the growth in communication alone is not sustainable any longer[3] On the other hand, recent advances in communication and informationtechnologies have fueled a plethora of innovations in various areas, includingnetworking, caching and computing, which have the potential to profoundlyimpact our society via the development of smart homes, smart transportation,smart cities[4], etc

re-Therefore, the integration of networking, caching and computing into onesystem becomes a natural trend [5,6] By incorporating caching functionalityinto the network, the system can provide native support for highly scalableand efficient content retrieval, and meanwhile, duplicate content transmissionswithin the network can be reduced significantly As a consequence, mobility,flexibility and security of the network can be considerably improved [7, 8]

On the other hand, the incorporation of computing functionality endows thenetwork with powerful capabilities of data processing, hence enabling the ex-ecution of computationally intensive applications within the network By of-floading mobile devices’ computation tasks (entirely or partially) to resource-rich computing infrastructures in the vicinity or in remote clouds, the taskexecution time can be considerably reduced, and the local resources of mobile

experience of mobile users [9] Moreover, networking, caching and computingfunctionalities can complement and reinforce each other by interactions Forinstance, some of the computation results can be cached for future use, thusalleviating the backhaul workload On the other hand, some cached contentcan be transcoded into other versions to better suit specific user demands,thus economizing storage spaces and maximizing the utilization of caching.Despite the potential vision of integrated networking, caching and com-puting systems, a number of significant research challenges remain to be ad-dressed before widespread application of the integrated systems, including thelatency requirement, bandwidth constraints, interfaces, mobility management,resource and architecture tradeoffs, convergence as well as non-technical issuessuch as governance regulations, etc Particularly, the resource allocation issue,

in which the management and allocation of three types of resources should bejointly considered to effectively and efficiently serve user requirements, is espe-cially challenging In addition, non-trivial security challenges are induced by alarge number of intelligent devices/nodes with self adaptation/context aware-ness capabilities in integrated systems These challenges need to be broadlytackled through comprehensive research efforts

In this chapter, we present a brief overview on the history and current search progress of networking, computing and caching technologies in wirelesssystems

The communication and networking technologies of the fifth generation (5G)wireless telecommunication networks are being standardized at full speedalong the time line of International Mobile Telecommunications for 2020(IMT-2020), which is proposed by the International Telecommunication Union(ITU) [10] Consequently, unanimous agreement on the prospect of a new,backward-incompatible radio access technology is emerging in the industry.Significant improvements on data rate and latency performance will bepowered by logical network slices [11,12] in the 5G communication environ-ment Furthermore, network flexibility will be greatly promoted by dynamicnetwork slices, therefore providing the possibility of the emergence of a va-riety of new network services and applications Due to the fact that the keyproblems of network slices are how to properly slice the network and at whatgranularity [11], it is not difficult to predict that a number of already exist-ing technologies, such as software defined networking (SDN) [13] and networkfunctions virtualization (NFV) [14], will be taken into consideration

SDN can enable the separation between the data plane and the controlplane, therefore realizing the independence of the data plane capacity fromthe control plane resource, which means high data rates can be achieved with-out incurring overhead upon the control plane [15, 16, 17] Meanwhile, theseparation of the data and control planes endows high programmability, low-complexity management, convenient configuration and easy troubleshooting

to the network [18] The controller in the control plane and the devices in thedata plane are bridged by a well-defined application programming interface(API), of which a well known example is OpenFlow [19] The instructions of acontroller to the devices in the data plane are transmitted through flow tables,each of which defines a subset of the traffic and the corresponding action Due

to the advantages described above, SDN is considered promising on providingprogrammable network control in both wireless and wired networks [20, 21].NFV presents the technology that decouples the services from the net-work infrastructures that provide them, therefore maximizing the utilization

of network infrastructure by offering the possibility that services from ferent service providers can share the same network infrastructure [14, 22].Furthermore, easy migration of new technology in conjunction with legacytechnologies in the same network can be realized by isolating part of the net-work infrastructure [14]

dif-On the other hand, heterogeneous network (HetNet) has been recentlyproposed as a promising network architecture to improve network coverageand link capacity [23, 24] By employing different sizes of small cells andvarious radio access technologies in one macro cell, energy efficiency (EE)and spectral efficiency (SE) are significantly enhanced in scenarios such asshopping malls, stadiums and subway stations [24,25,26,27] But ultra-densesmall cell deployment coexisting with legacy macro cells raise the problem

of mutual interference, which calls for efficient radio resource allocation andmanagement technologies [28,29,30]

Meanwhile, the Cloud Radio Access Network (C-RAN) is rising as apromising solution for reducing inter-cell interference, CAPital EXpenditure(CapEx) and OPerating EXpenditure (OpEx) [31] Based on network central-ization and virtualization, C-RAN is dedicated to promoting energy efficiency,coverage and mobility performances, while reducing the expense of networkoperation and deployment [31] By pooling the baseband resources at theBaseBand Unit (BBU), which is located at a remote central office (not at thecell sites), C-RAN is able to provide the network operator with energy efficientoperations, statistical multiplexing gains and resource savings [32,33].Massive multiple-input multiple-output (MIMO) technology was identified

as a key 5G enabler by the European 5G project METIS in its final deliverable[34] In a massive MIMO scenario, K pieces of user equipment (UE) are served

on the same time-frequency resource by a base station (BS) equipped with Mantennas, where M ≫ K The deployment of a large number of antennas at the

BS leads to a particular propagation scenario named favorable propagation,where the wireless channel becomes near-deterministic since the BS-to-UE

fields like 3D MIMO, hybrid beamforming (BF) and understanding of theasymptotic behavior of random matrices suggest that massive MIMO hasthe potential to bring unprecedented gains in terms of spectrum and energyefficiencies and robustness to hardware failures [34].

With the severe shortage in the currently utilized spectrum (sub-6 GHz),the utilization of higher frequency bands, e.g., millimeter-wave (mmWave),becomes a consensus of the communication community [36] Working on 3–

300 GHz bands, mmWave communications assisted by smart antenna arrayscan track and transmit signals to high-speed targets over a long distance [37].Since the physical antenna array size can be greatly reduced due to a de-crease in wavelength, mmWave communications are appealing to large-scaleantenna systems [38] On the other hand, the detrimental effects of the highpropagation loss in mmWave communications can be neatly compensated bythe achievable high beamforming gains with massive MIMO [36, 39] There-fore, mmWave communication assisted by massive MIMO is envisaged as apromising solution for future telecommunications

Another way to deal with spectrum scarcity is to increase the utilizationefficiency of the spectrum Allowing two pieces of UE to perform direct datatransmissions, device-to-device (D2D) communications enable a flexible reuse

of radio resources, and therefore improve the spectral efficiency and ease corenetwork data processing workloads [40,41] Moreover, the combination of D2Dcommunications with information-centric networking (ICN) is worth consid-ering Despite the small-sized storage of UE, the ubiquitous caching capabilityresiding in UE should not be neglected, due to the UE’ pervasive distributionand ever-increasing storage sizes [42]

As described above, the spectral efficiency (SE) of wireless communicationradio access networks has been increased greatly by ultra-dense small cell de-ployment However, this has brought up another issue: the backhaul may be-come the bottleneck of the wireless communication system due to the tremen-dous and ever increasing amount of data being exchanged within the network[43] On the other hand, building enough high speed backhauls linking thecore network and the small cells which are growing in number could be ex-ceptionally expensive Being stimulated by this predicament, research effortshave been dedicated to caching solutions in wireless communication systems[44] By storing Internet contents at infrastructures in radio access networks,such as base stations (BSs) and mobile edge computing (MEC) servers [45],caching solutions enable the reuse of cached contents and the alleviation ofbackhaul usage Therefore, the problem has been transferred by caching solu-tions from intensive demands on backhaul connections to caching capability

of the network

However, the original intention of Internet protocols is providing direct

connections between clients and servers, while the caching paradigm calls for

a usage pattern based on distribution and retrieval of content This diction has led to a degradation of scalability, availability and mobility [46]

contra-To address this issue, Content Delivery Networks (CDNs) [48] and Peer (P2P) networks [49] have been proposed in application layers, as firstattempts to confer content distribution and retrieval capabilities to networks

Peer-to-by utilizing the current storage and processing network infrastructures.CDNs employ clusters of servers among the Internet infrastructures andserve the UE with the replications of content that have been cached in thoseservers The content requests generated by UE are transmitted through theDomain Name Servers (DNSs) of the CDN to the nearest CDN servers thathold the requested content, in order to minimize latency The decision on whichserver is chosen to store the replication of content is made upon a constantmonitoring and load balancing of data traffic in the network [50]

P2P networks rely on the storage and forwarding of replications of content

by UE Each UE can act as a caching server in P2P networks [50, 51] InP2P networks, the sources of content are called peers Instead of caching acomplete content, each peer may store only a part of the content, which iscalled a chunk Thus, the content request of a UE is resolved and directed

to several peers by a directory server Each peer will provide a part of therequested content upon receipt of the request

Although CDN and P2P do give a solution for content distribution andretrieval, due to the fact that they solely operate on application layers andthe commercial and technological boundaries that they are confined to, theperformances of these two techniques are not ideal enough to fulfil the demands

on network caching services [50]

Serving as an alternative to CDN and P2P networking, Centric Networking (ICN) emphasizes information dissemination and contentretrieval by operating a common protocol in a network layer, which can uti-lize current storage and processing network infrastructures to cache content

Information-In general, depending on the location of caches, caching approaches of ICNcan be categorized as on-path caching and off-path caching [52, 53] On-pathcaching concerns the delivery paths when considering caching strategies, andhence are usually aggregated with the forwarding mechanisms of ICN Onthe other hand, off-path caching solely focuses on the storage and delivery ofcontent, regardless of the delivery path

As the prevalence of smartphones is dramatically growing, new mobile tions, such as face recognition, natural language processing, augmented reality,etc [1] are emerging This leads to a constantly increasing demand on compu-tational capability However, due to size and battery life constraints, mobiledevices tend to fail in fulfilling this demand On the other hand, powered bynetwork slicing, SDN and NFV, cloud computing (CC) [54] functionalities are

applica-of mobile cloud computing (MCC) [55], which leverages the rich tion resources in the cloud for user computation task execution, by enablingcomputation offloading through wireless cellular networks After computa-tion offloading, the cloud server will create a clone for each piece of userequipment (UE) individually, then the computation task of the UE will beperformed by the clone on behalf of that UE Along with the migration

computa-of computation tasks from UE to a resourceful cloud, we are witnessing anetworking paradigm transformation from connection-oriented networking tocontent-oriented networking, which stresses data processing, storage and re-trieval capability, rather than the previous criterion that solely focuses onconnectivity

Nevertheless, due to the fact that the cloud is usually distant from mobiledevices, the low-latency requirements of some latency-sensitive (real-time) ap-plications may not be fulfilled by cloud computing Moreover, migration of alarge amount of computation tasks over a long distance is sometimes infeasibleand uneconomical To tackle this issue, fog computing [56,57,47,59] has beenproposed to provide UE with proximity to resourceful computation servers.The terminology fog (From cOre to edGe) computing was first coined in 2012

by Cisco [60] It is a distributed computing paradigm in which network entitieswith different computation and storage abilities and various hierarchical levelsare placed within a short distance from the cellular wireless access network,connecting user devices to the cloud or Internet It is worth noting that fogcomputing is not a replacement but a complement of cloud computing, due

to the fact that the gist of fog computing is providing low-latency services tomeet the demands of real-time applications, such as smart traffic monitoring,live streaming, etc However, when the applications requiring a tremendousamount of computation or permanent storage are concerned, the fog comput-ing infrastructures are only acting as gateways or routers for data redirection

to the cloud computing framework [57]

The problems in wireless networks such as congestion of Internet tions, low-bandwidth and infeasibility of real-time applications can hardly besolved by simply and blindly boosting the underlying network bandwidth ca-pacity By analyzing real data collected from three hotels in Italy, the work in[59] shows that fog computing is competent in alleviating the effects of thoseproblems This analysis demonstrates that by deploying fog nodes and cor-responding applications at the network edge, fog computing can proactivelycache and manage up to 28.89 percent of the total traffic, which cannot bemanaged by conventional networking and caching approaches Moreover, byimposing selective bandwidth limitations on specific connected devices, thefog nodes enable a dynamic and flexible management of the available band-width, bringing benefits to wireless network in terms of resource optimizationand performance [59] This real data based analysis can serve as a prelimi-nary validation of the feasibility and significance of the incorporation of fogcomputing into wireless networks

connec-Also intended to solve the long-latency problem of mobile cloud puting, cloudlet-based mobile cloud computing is proposed [61] Instead ofutilizing the computation resource in a remote cloud, cloudlet-based mobilecloud computing can reduce data transmission delay by deploying comput-ing servers/clusters within one-hop WiFi wireless access However, there aretwo drawbacks associated with cloudlet based mobile cloud computing: First,cloudlet-based mobile cloud computing can hardly guarantee pervasive ser-vice due to the limited coverage of the WiFi access network which is usuallyemployed in indoor environments Second, due to the constraints on infras-tructure size, the servers of cloudlet-based mobile cloud computing typicallyprovide a small or medium amount of computation resources, which can hardlymeet the computation requirement in the case of a large number of UE.

com-To serve as a complement of cloudlet-based mobile cloud computing, anew MCC paradigm similar to fog computing, named mobile edge computing(MEC), is proposed [45,62] Being placed at the edge of radio access networks,MEC servers can provide sufficient computation resources in physical proxim-ity to UE, which guarantees the fulfilment of the demand of fast interactiveresponse by low-latency connections Therefore, mobile edge computing is en-visioned as a promising technique to offer agile and ubiquitous computationaugmenting services for mobile devices, by conferring considerable computa-tional capability to mobile communication networks [14]

In this chapter, we first define the integrated system of networking, cachingand computing in question, then we discuss the motivations of the integration,followed by the requirements of the integration of networking, caching andcomputing

computing?

In order to give a clear description of integrated networking, caching andcomputing, we first give the definitions of networking, caching and computingconsidered here respectively

 Networking

The networking vector discussed in this chapter pertains to its bility to deliver data through networks with certain bandwidth andpower The measurement of this capability is usually data rate, whoseunit is bits per second The relationship among data rate, bandwidthand signal to noise power ratio is well demonstrated by Shannon’scapacity formula

capa-The caching vector considered in wireless communication systems tains to its ability to store a certain amount of data at the infrastruc-tures in the network The typical measurement of caching capability

per-is the size of stored information, whose unit per-is bytes Without ing the information flow in network, caching alleviates backhaul andenhances the capability of information delivery over a long term

The computing vector under consideration in this chapter pertains toits capability to perform algebraic or logical operations over the infor-mation flows within the network In contrary to the networking andcaching vectors, the operation of the computing vector is intended toalter the information flow in the network One of the typical measure-ments of computing capability is the number of information flows ornodes involved in the computing operation, which is called the degree

of computing (DoC) [3]

Given the definitions of networking, caching and computing in wireless

or mobile systems, we next move on to the discussion of the topic of thischapter, integrated networking, caching and computing In this chapter, weconsider integrated frameworks and technologies in wireless or mobile commu-nication systems that take into consideration at least two aspects of network-ing, caching and computing as defined above, aiming at providing coherentand sustainable services to fulfil various mobile network user demands, such

as high-data-rate communication, low-latency information retrieval and tensive computation resources In other words, instead of solely focusing onthe networking vector, we consider solutions concerning three network vec-tors, namely, networking, caching and computing, as shown in Figure 1.1.The acronyms IoT, AR and SMS inFigure 1.1stand for Internet of Things,Augmented Reality and Short Message Service, respectively

caching and computing?

1.2.2.1 The growth of networking alone is not sustainable

Since the first application of commercial mobile cellular services around 1983,excessive research endeavors have been dedicated to the improvement of thecommunication capacity of mobile communication systems [3] It is unani-mously granted that in the early days of mobile communication systems, whenthe main functionality of the system was voice and texting services, improve-ment of throughput, latency and link capacity could significantly enhance theuser experience, which in turn yielded greater profits for network operators.However, as years passed, the mobile communication systems have undergone

Figure 1.1: Schematic graph of integrated networking, caching and computing.

an evolution, in which their functionalities have grown Especially in recentyears, along with the growing prevalence of smartphones, new applications areemerging constantly

Due to the fact that many multimedia applications rely on a tremendousamount of data, the demands of mobile network users on powerful storageservices are rapidly increasing According to the report of ARCchart [64],with the radically increasing camera resolution of mobile devices and the evenfaster growing size of data generated by such hardware advances, demands onmobile cloud storage are undergoing an exponential growth To meet this userrequirement, mobile cloud storage has risen and taken the place of the mostwidely used cloud mobile media (CMM) service, which is mainly provided byGoogle, Apple, Dropbox and Amazon today [1] This service enables mobilenetwork users to store video, music and other files in the cloud, to access filesthrough any devices regardless of the source of the data, and to synchronizedata in multiple devices In order to employ a mobile cloud storage service

on a large scale, it is imperative to ensure high integrity and availability ofdata, as well as user privacy and content security It is clear that large scaledeployment of a mobile cloud storage service will cause a significant increase

in mobile data traffic, which in turn puts a demand on the integration of thestorage functionality with ubiquitous and high-data-rate wireless networkingfunctionality

On the other hand, audio and video streaming based service, which ically needs excessive computation resources, is becoming one of the majorservices in mobile communication systems [1] According to the technical re-port of CISCO [65], by the end of 2017, video traffic is expected to take

typ-zettabytes ∗ However, the very limited computation resources provided bymobile devices can hardly meet the requirement raised by tasks such as en-coding, transcoding and transrating in this category of service Therefore, tosolve this dilemma, the incorporation of powerful computing technologies likeMCC and MEC into mobile communication systems becomes inevitable Byleveraging services of MCC and MEC, not only can the time consumption

of computing be reduced, but also computing and backhaul costs can be duced by caching popular videos through the caching functionality of MCCand MEC Furthermore, the elasticity of computation resources provided byMCC/MEC servers is suitable for handling volatile peak demands in a cost-effective pattern

re-In summary, the sole emphasis on networking capabilities such as tivity and spectral efficiency can hardly fulfill user demands on the diversity

connec-of services, especially services that rely on intensive computation resourcesand enormous storage capabilities, and thus the growth of networking alone

is unsustainable The integration of networking, caching and computing isindispensable to achieve sustainability of mobile communication systems

1.2.2.2 The benefits brought by the integration of networking,

caching and computing

The incorporation of caching functionality into mobile communication systemsenables information distribution and high-speed retrieval within the network.In-network caching and multicast mechanisms facilitate the transformation ofdata delivery pattern from direct server-client link to an information-centricparadigm, thus enhancing the efficiency and timeliness of information delivery

to users [52] Furthermore, network traffic reduction can also be realized byemploying in-network caching in mobile communication systems According

to the trace-driven analysis in [66], with an in-network caching mechanism,three or more hops for 30% of the data packets in the network can be saved,

or equivalently 2.02 hops on average can be saved Moreover, up to 11% ofusers’ requests for data can be retained within the requester domain

On the other hand, the incorporation of computing functionality into bile communication systems brings the benefits of access to the rich computa-tion and storage resources in the cloud or the network edge, which enables theutilization of cutting edge multimedia techniques that are powered by muchmore intensive storage and computation capabilities than what the mobiledevices can provide, thus offering richer and finer multimedia services thanwhat primitive mobile applications can provide Moreover, due to the fact thatmobile applications enabled by MCC/MEC technologies can exploit not onlythe resources in the cloud or network edge, but also the unique resources inmobile devices, such as location information and camera data, these mobile

mo-∗ 1 zettabyte=10 21

bytes.

applications are apparently more efficient and richer in experience than PCapplications.

The integration framework makes it possible for users to access media andinformation through any platform, any device and any network According

to Juniper Research, the revenues of consumer cloud mobility services, whoseinitial prototypes are cloud based video and music caching and downloadingservices like Amazon’s Cloud Drive and Apple’s iCloud, reached 6.5 billiondollars per year by 2016 [67]

caching and computing

In order to implement the integration of networking, caching and computing,several requirements need to be met

1.2.3.1 Coexistence

In an integrated framework, it is important for the caching and computinginfrastructures to coexist with the mobile communication infrastructures, toguarantee the seamless cooperation of the networking, caching and computingfunctionalities Actually, coexistence is the prerequisite of the realization ofintegrated networking, caching and computing systems [10]

The infrastructure type, topology, QoS requirement, service type and rity level are all different across these three functional vectors, and when theyare integrated into one framework, they still need to bear all these differentproperties

secu-1.2.3.2 Flexibility

A certain degree of freedom in these three functional vectors is necessary Theflexibility depends on the degree of the integration of the three vectors [3] Ahigh level of integration may provide good multiplexing of the three vectors,seamlessness of cooperation between different functionalities and simplicity ofimplementation, while reducing the flexibility of resource customization in theintegrated system A low level of integration can lead to the reverse effects

1.2.3.3 Manageability and programmability

The configuration, deployment and various resource allocation of the grated framework bring up the requirements of manageability and programma-bility [3] Only with a relatively high degree of manageability and programma-bility can the admission, maintenance and resource scheduling of the inte-grated system be viable To realize programmability, the system needs toprovide effective programming language, feasible interfaces and a secure pro-gramming environment with a considerable level of flexibility

inte-Different functional vectors, networking, caching and computing, typically erate in different network paradigms, and a heterogeneous network structurecan facilitate the coexistence of different networking access manners [10].

op-1.2.3.5 Scalability

Since the number of mobile users and the complexity of mobile applicationsare increasing as time goes by, the integrated system of networking, cachingand computing should have the capability to support a fluctuating number of

UE and complexity of tasks

1.2.3.6 Stability and convergence

The integration of networking, caching and computing should be stable enough

to overcome errors, misconfiguration and other unstable situations Also, anyalgorithm concerning the allocation of multiple resources should have a goodconvergence property

1.2.3.7 Mobility

The integrated system should be able to support not only the geographicalmobility of mobile users, but also the logical mobility of caching and comput-ing functional vectors within the system

1.2.3.8 Backward compatibility

Integrated systems should be able to exploit the existing communication work infrastructures up to the hilt The caching and computing functionali-ties should be realized on the basis of utilizing existing networking resources.Therefore, from the economic point of view, the backward compatibility withlegacy communication systems is greatly beneficial and vital This means thatthe compatibilities on both interfaces and protocols should be taken into con-sideration when designing integrated systems

In this chapter, we summarize the typical frameworks (architectures) of grated systems of networking, caching and computing The frameworks pre-sented are based on existing studies, and reflect the basic ideas, mechanismsand components of integrated systems Unfortunately, this brief summary maynot cover all of the proposed architectures in the literature, due to the factthat each proposed architecture has its unique original intention, idea of designand working environment

inte-Figure 1.2: Networking-caching framework [8].

In this subsection, we will stress on the study of [68], due to the fact that

it provides a rough summary of various typical caching approaches withinone framework Without losing generality, we will also give a quick glance atframeworks proposed by other studies

The framework proposed in [68] intends to provide a distributed cachingand content delivery solution, taking into consideration diverse storage, con-tent popularity distribution and user device mobility In this framework, thecommunication and caching capabilities of small cell base stations (SBSs) anddevice-to-device (D2D) enabled mechanisms are jointly utilized to support

a multilayer content caching and delivery framework Both SBSs and userdevices have the ability to store and transmit content

The multilayer caching-networking framework is illustrated inFigure 1.2.Multiple small cells are deployed in one macro cell, and all the SBSs are con-nected to the macro cell base station (MBS) through backhaul links TheMBS is connected to a gateway of the core network via a high-speed interface.Each individual SBS and UE is equipped with a caching device to performthe caching function According to current standards of the Third GenerationPartnership Project (3GPP), X2 interfaces are solely intended for signalingbetween SBSs, and therefore cached content sharing between SBSs is not per-mitted However, by utilizing inter-cell D2D communications, content cached

in different small cells can be shared with each other

As shown inFigure 1.2, this framework consists of the following five cachedcontent delivery approaches

Content requested by UE can be retrieved through D2D links from neighboring

UE A high density of UE can contribute to a high probability of successfulretrieval of requested content Moreover, the implementation of inter small cellD2D communications can further facilitate the utilization of content cached

in adjacent small cells

D2D communication links can serve as a relay for content delivery from othersources to the destination UE This approach enables a broader range ofcaching and delivery

If multiple UE held the content requested by another UE, they can atively transmit the content to the destination UE through a D2D multiple-input multiple-output (MIMO) technology to accelerate the transmission pro-cess Moreover, if video coding techniques such as multiple description codingand scalable video coding are taken into consideration during video data de-liveries, more benefits may be gained

If the requested content was cached in the associated SBS, the SBS can mit it directly to the destination UE Despite the interference incurred duringthe retrieval and transmission processes, the latency can be lower than mul-tihop transmission approaches

If the content requested by a certain UE was stored neither in its nearby UEnor in its associated SBS, it may retrieve and demand the content from an SBS

in a neighboring small cell In this case, the associated SBS will require thedemanded content from neighboring SBSs through virtual links, which meansthese two SBSs are able to communicate with each other via their respectivebackhauls to the MBS

Three function modules, namely, REQProxy, ContentServer, and Server, are designed in the application layer to enable the above five con-tent delivery approaches REQProxy recognizes content requests from UEand transmits the requests to ContentServer ContentServer performs con-tent retrieving and forwarding GateServers are embedded in SBSs, and theysend content requests that cannot be handled locally to other SBSs or remoteservers

Gate-The signaling interaction procedures of this framework are shown in

Figure 1.3: Signaling interaction of the networking-caching framework [68].

to demonstrate the signaling interaction procedures

Cooperative D2D delivery (Fig 1.2c) When any content is needed by acertain UE, the Content REQ (content request) is sent by the UE to nearby

UE Then each individual receiver’s REQProxy will contact its own tentServer to check whether the requested content is cached by itself If so,the REQProxy of the receiver will respond to the REQProxy of the requesterwith a RESP ACK (acknowledgment) After that, the content transmissionlink will be established with another two packets, Establish Connection andConnection ACK This signaling interaction procedure is also applicable inthe process of plain D2D delivery (Fig 1.2a) and Direst SBS delivery (Fig.1.2d), where the associated SBS responds to the requester with RESP ACK.Cooperative SBS delivery (Fig 1.2e) In this approach, the requester UEsends a Content REQ to its associated SBS, then the SBS retransmits thisContent REQ to the MBS, through GateServer The virtual link between theSBS holding the requested content and the requester’s associated SBS will

Con-be established Then the content will Con-be transmitted to and cached at theassociated SBS and eventually delivered to the requester UE

Multihop delivery via D2D relay (Fig 1.2b) In this approach, the relay

UE needs to establish a link with its associated SBS When the requestedcontent is ready to be transmitted, the SBS will send a connection ACK tothe relay UE in order to establish transmission connection

Figure 1.4: The information-centric networking [8].

An important feature of 5G mobile networking could be recognizing thecontent as the first-class citizen in the network [8] To fulfil this requirement,the conventional content delivery networking (CDN) [69,70] has distributedfurther, incorporating content-aware and information-centric caching tech-nologies, and forming the so-called information-centric networking (ICN) [71],which has the significant advantages of natively facilitating efficient and scal-able content retrieval and enhancing the mobility and security of the network

As an important native feature of ICN, in-network caching has the capability

of significantly reducing duplicate data transmissions in the network based air caching technology is considered a promising candidate to supportthe implementation of the SDN-based 5G mobile network [71]

ICN-The study of [8] proposes an information-centric wireless network (ICN)virtualization framework, which involves content-level slicing functionality Asshown inFigure 1.4, in contrast with Internet Protocol (IP), which adopts ahost-based conversation model (i.e., establish and maintain connection be-tween hosts before actual data delivery) and follows a source-driven data de-livery approach (i.e., path is initiated by sender to receiver), ICN follows areceiver-driven communication principle (i.e., path is initiated by receiver tosender), and the data transmission follows the reverse way As can be seen in

it and disseminate the content to other UE Instead of the reachability ofthe hosts and the maintenance of the connectivity between hosts, the mainconcern of ICN is retrieval and dissemination of information

The authors of [8] present a content-level slicing in their proposed tualized ICN framework, which can be considered an extension of dynamiccontent access and sharing The physical content (cache) is sliced throughspace multiplexing, time multiplexing and so forth, then allocated to differ-ent service providers [72] In this approach, several services share a number

vir-of physical contents cached in the network, and each content is sliced into

a number of virtual contents Each service can utilize one or several of thevirtual contents without the knowledge of other virtual contents The number

of virtual copies of the physical contents can depend on the popularity of thecontent, i.e., more popular physical content is sliced into more virtual contents

Conceptually speaking, content-level slicing is a realization of caching sharingand dynamic access in a virtualized networking environment.

Not only the reduction of network data traffic but also the improvement

of quality of experience can be realized by moving the frequently requestedcontent toward the network edge closer to UE [71] The study of [73] pro-poses a framework of caching popular video clips in the RAN The concept

of FemtoCaching is proposed in [74], and the video content caching and livery approach with the facilitation of distributed caching helpers in fem-tocell networks is discussed The research conducted in [71] studies currentcaching technologies and discusses caching in 5G networking, including ra-dio access network caching and evolved packet core network caching Then anedge caching approach (on the edge of RAN) based on information-centric net-working is designed The paper discusses architecture design issues on cachingwithin evolved packet core (EPC), caching at RAN and content-centric net-working (CCN) based caching

de-As the popularity of social media is increasing dramatically, online socialnetworks (OSN) are starting a new revolution in how people communicate andshare information In centralized OSN, the network has a central repository

of all user data and imposes control regarding how user data will be accessed.This form of architecture has the advantage of efficient dissemination of socialupdates However, users do not have control over their personal data, andare therefore constantly exposed to potential privacy violations To addressthis issue, the architecture of distributed online social networks (DOSN) hasbeen proposed [75, 76] DOSN does not have a central repository and henceguarantees the user privacy, but faces the problem of inefficient dissemination

of social updates The work in [75] introduces Social Caches as a way to leviate the peer-to-peer traffic in DOSNs, and the social caches are supposed

al-to act as local bridges among friends al-to provide efficient information delivery.Furthermore, a social communication model called Social Butterfly has beenproposed to utilize social caches However, the selection of social caches de-mands full knowledge of the whole social graph Therefore the authors in [76]propose a number of distributed social cache selection algorithms

Due to the fact that stable D2D links can be easily maintained among ple who have close social relationships, social-aware device-to-device (D2D)communication is proposed to complement the online social network com-munication The work in [77] proposes a fog radio access network (FRAN)supported D2D communication architecture, in which the UE can downloadfiles either from a centralized baseband unit (BBU) pool through distributedremote radio heads (RRHs), or from other UE through D2D links In [78],

peo-a context-peo-awpeo-are frpeo-amework for optimizing resource peo-allocpeo-ation in smpeo-all cellnetworks with D2D communication is proposed Both the wireless and so-cial context of wireless users are utilized to optimize resource allocation andimprove traffic offload

Figure 1.5: Cloud mobile media framework, demonstrating data and controlflows [1].

In this subsection, we will consider the frameworks of Cloud Mobile Media(mobile cloud computing) and mobile edge computing as two representa-tives to give a general profile of the architecture of the computing-networkingparadigm

1.3.2.1 Cloud mobile media

The overall architecture of cloud mobile media (CMM) is depicted inFigure1.5, including end to end control and data flows between Internet cloud serversand mobile devices [1] Enabled by cloud mobile media, applications on mo-bile devices are able to utilize intensive and elastic computing and cachingresources in remote clouds (typically located in the Internet), including pri-vate, public and federated (hybrid) cloud types The user interfaces of CMMapplications used for command input are provided by user devices, includingtouchscreens, gestures, voices and texts The subsequent control commandsare then transmitted uplink via cellular radio access networks (RAN) to gate-ways located in core networks (CN), and eventually to the serving clouds.After that, the clouds perform data processing or content retrieval using com-puting or caching resources in cloud servers, in accordance with user controlcommands Then the results are sent back to user devices downlink through

CN and RAN

Please note that most of the applications follow the data and control flowsshown inFigure 1.5, but there exist some exceptions that slightly deviate fromthem For instance, the control flows in cloud-based media analytics are notalways initiated by mobiles devices, and this type of application may collectdata from both the cloud and the mobile devices in order to perform analyticsfor other types of CMM applications

Figure 1.6: MEC server deployment scenarios [45].

1.3.2.2 Mobile edge computing

As described in previous sections of this chapter, MEC serves as a complement

to mobile cloud computing, and the intention of MEC is reducing the highdata transmission latency of MCC by providing proximity to computationresources at the edge of a radio access network The basic mechanism of MEC

is quite similar to that of MCC They both allow UE to offload computationtasks to servers that execute the tasks on behalf of the UE, then the resultsare returned to the UE

MEC servers can be deployed at various locations within the edge of dio access networks According to the report of [45], the MEC server can bedeployed at the macro base station (eNB) site, at the multi-technology cell ag-gregation site, or at the radio network controller (RNC) site The deploymentscenarios are depicted in Figure 1.6 The multi-technology cell aggregationsite can be located either at an indoor environment in an enterprise, such as auniversity center or airport, or at an outdoor environment to control severallocal multi-technology access points providing coverage over certain publicpremises Direct delivery of fast, locally relevant services from base stationclusters to UE can be enabled by this option of deployment

In recent years big data has attracted more and more attention due to theincreasing diversity of mobile applications and the tremendous amount of datastreaming in mobile networks In this context, data management/processingand data caching have emerged as two key issues in this field The study

Figure 1.7: Caching-computing architecture [79].

of [79] proposes a caching-computing framework in which cached content ismoved from the cloud to the RAN edge, and a big data platform powered

by machine learning and big data analytics is deployed at the core site forcontent selection

As illustrated in Figure 1.7, the caching platform at small cell base tions (SBSs) and the computation and execution of the content predictionalgorithms at the core site are two key components of the proposed architec-ture A big data platform is employed at the core site to perform trackingand prediction of users’ demands and behavior for content selection decisions,and SBSs in the RAN are equipped with caching units to store the strategiccontent selected by the big data platform The reason for employing the bigdata platform is that the content cached in SBSs can hardly cover all theusers’ content demands, due to the limited storage capacity of the SBSs Inorder to yield optimal performance, it is important to select the most popu-lar content (most frequently requested content based on estimation) and todecide the content cache placement at specific small cells based on backhauls,rate requirements, content sizes and so forth

sta-The big data platform portrayed inFigure 1.7is used for sorting users’ datatraffic and extracting statistical information for proactive caching decisions

To achieve this goal, the following requirements should be met by the big dataplatform

Enormous amount of data processing in a short time period Making tive caching decisions requires the capability of processing a large amount ofdata and drawing intelligent insights in a short time

proac-Cleansing, parsing and formatting data In order to perform statisticalanalysis and machine learning, the data should be cleaned in advance Themalfunctioning, inappropriate and inconsistent packets involved in the rawdata should be eliminated before any type of data processing Then the rel-evant fields should be extracted from the raw data Finally, the parsed datashould be encoded accordingly

Data analysis A variety of data analysis technologies can be applied inthe analysis of the header and/or payload information of both control anddata planes The purpose of data analysis is finding the relationship between

Figure 1.8: Architecture of cloud content-centric mobile networking [80].

control/data packets and the requested content and then predicting temporal user behavior

spatial-Statistical analysis and visualizations The outcomes of the data analysiscan be stored and reused for further statistical analysis Moreover, the out-comes can be formatted as graphs or tables for enhancement of readabilityand ease of comprehension

In [80], the authors propose a cloud content-centric mobile networking(CCMN) architecture, which incorporates cloud computing into both corenetwork caching and radio access network caching In this architecture, theconventional content delivery network (CDN) migrates into the cloud, forming

a cloud CDN which leverages the merits of both cloud computing and CDN As

an application of cloud computing, the cloud CDN architecture inherits most

of the advantages of cloud computing, and moreover, it can offer sufficientelasticity for handling flash crowd traffic However, the content proximity to

UE and the coverage of cloud CDN are not as ideal as those of conventionalCDNs

The merging of cloud computing and RAN forms the so-called cloud RAN(C-RAN), whose most significant innovation is that the computationally inten-sive baseband processing tasks are executed by a centralized cloud basebandunit (BBU) pool, instead of the traditional distributed baseband processingdevices at the base station sites [81] By adding the content caching function-ality into conventional C-RAN, C-RAN with caching as a service (CaaS) isconsidered a promising approach to overcome the limitations of core networkcaching, due to the fact that the content is moved from a core network to

a RAN, and hence is much closer to UE In C-RAN with CaaS, the BBUpool can offer larger caching space than that of BSs, and the centralized BBUpool can facilitate solving the distributed content placement problem over ge-ographically scattered BSs Nevertheless, C-RAN with CaaS doesn’t do well

in fronthaul traffic mitigation

The overall architecture of the cloud content-centric mobile networking(CCMN) is depicted in Figure 1.8, which shows the assistant role of cloudcomputing in both core network caching and RAN caching

A similar framework called fog RAN (F-RAN), combining edge cachingand C-RAN, has been recently advocated [82,83,84] In F-RAN, edge nodesare equipped with caching capabilities, and simultaneously are controllable by

efficiency due to cooperative cloud-based transmission, while the latency isnot ideal because of fronthaul transmission On the other hand, edge cachingcan provide low-latency deliveries of popular content, but suffers from poorinterference mitigation performance due to decentralized baseband processing

at the edge nodes By leveraging features of both edge caching and C-RAN,F-RAN can benefit from both low-latency deliveries of content and centralizedbaseband processing

In this subsection, we generally consider two categories of computing frameworks, namely, networking-caching-computing convergenceand networking and computing assisted caching The former refers to theframework in which networking, caching and computing functionalities coa-lesce with each other to form a comprehensive structure, providing UE withversatile services The latter means in this type of frameworks networking andcomputing capabilities of the network elements are functioning to provide sup-ports to caching functionality, improving the quality of caching services of thenetworks

networking-caching-1.3.4.1 Networking-caching-computing convergence

The synergy of networking, caching and computing is not a trivial step First,different resource accesses require unique communication protocols, and there-fore the convergence of these three types of resource accesses poses a require-ment on new protocol design Second, interactions between different types ofresources need appropriate interfaces on network devices, which means re-design of network devices is necessary

The authors in [5] propose a framework incorporating networking, cachingand computing functionalities under the control and management of an SDNcontrol plane, aiming at providing energy-efficient information retrieval andcomputing services in green wireless networks A software-defined approach

is employed in the integrated framework, in order to leverage the separation

of the control plane and the data plane in SDN, which help guarantee bility of the solution Three types of radio access networks are considered inthis chapter, namely, cellular networks, wireless local area networks (WLANs)and worldwide interoperability for microwave access (WiMAX) Each networknode is equipped with caching and computing capabilities, and is under thecontrol of a controller, which is in charge of the topology of the whole wire-less network and the packet forwarding strategies Thus, this framework canprovide UE with caching, computing, as well as communication services

flexi-1.3.4.2 Networking and computing assisted caching

Caching systems face three primary issues: where to cache, what and how tocache, and how to retrieve To answer the question where to cache, identifica-tion of eligible candidates for caching the content is needed In other words,

it is necessary to clarify the criteria of a node being an information hub inthe network Following that, the question what to cache with respect to thepopularity and availability of the content is posed It is also imperative to de-cide which nodes should keep which content for redundant caching avoidance.After the content is appropriately cached in the network, the final questionhow to retrieve requires an answer

In order to accommodate the dramatically increasing content requests erated by a huge number of UE in geographical proximity in vehicular net-works, the study in [85] advocates the adoption of a networking and computingassisted caching paradigm in information-centric networking architecture, forICN can provide the decoupling of content providers and consumers, and in-network caching functionality is intrinsically supported by ICN Therefore,the paper proposes a socially aware vehicular information-centric networkingmodel, in which mobile nodes such as vehicles are equipped with commu-nication, caching and computing capabilities, and they are responsible forcomputing their eligibility for caching and delivering content to other nodes(i.e., the relative importance of vehicles in the network, with respect to userinterests, spatial-temporal availability and neighborhood connectivity) Fur-thermore, each node also bears the mission of computing its eligibility for be-ing an information hub in the network This has addressed the where to cachequestion For the what and how to cache question, the decision is made on thecomputation results regarding content popularity, availability and timeliness,taking into consideration cooperative caching schemes As for the how to re-trieve question, the paper presents a social content distribution protocol forvehicles to relay and retrieve cached content, in order to enhance informationreachability

tech-to one MEC server, and they are all connected tech-to the Internet through thecore network A cache (storage facility) is deployed at each virtual BS.Joe finds an interesting monument when he is visiting Ottawa With hisaugmented reality (AR) glasses, Joe notices that there is an introductory video

Figure 1.9: A use case of the integrated system of networking, caching andcomputing.

about this monument Being interested in this video, he generates a videocontent request via his AR glasses to a virtual BS According to the description

of the video content and the information about the user, the virtual BS willcheck whether its associated cache has the requested content If yes, the virtual

BS will further examine if the version of its cached video content matches theuser’s device If still yes, the virtual BS will directly send the requested content

to the user from the cache If no, the virtual BS will extract the current videocontent and initiate a computation task according to the size of the contentincluding the input data, codes and parameters, as well as the number ofCPU cycles that is needed to accomplish the computing/transcoding task.Then the virtual BS will transmit the task to the MEC server to executethe computation After the computation is finished, the virtual BS sends thetranscoded video content to the user If the virtual BS cannot find any version

of the requested video content in the cache, the virtual BS has to retrievethe content from the Internet, and this will inevitably consume some of thebackhaul resources Upon the arrival of the new video content at the virtual

BS, the virtual BS can choose whether to store the content in its cache or not.The procedure is shown inFigure 1.10

Figure 1.10: The processing procedure of the use case

The above procedure is for downlink streaming video Similarly, the dure for uplink streaming video can be considered as follows When Joe finds

proce-an interesting monument, he takes a video of it, puts some comments on it,and streams it to a remote server via a virtual BS The virtual BS extracts thevideo content, and estimates the possibility that other people want to see thisvideo If the possibility is high, the virtual BS will cache this video locally, sothat other people who are interested in this video can obtain this video fromits cache directly, instead of requesting it from the remote server As thus,backhaul resources can be saved and propagation latency can be reduced Ifthe cached version does not match the new user’s device, transcoding will beperformed by the MEC server

Although the above use case focuses on tourism services, similar use caseswith similar requirements can be derived for other services and applications,such as transportation, education, healthcare, etc

[3] H Liu, Z Chen, and L Qian, “The three primary colors of mobile tems,” IEEE Commun Mag., vol 54, no 9, pp 15–21, Sept 2016.[4] Y He, F R Yu, N Zhao, V C M Leung, and H Yin, “Softwaredefinednetworks with mobile edge computing and caching for smart cities: A bigdata deep reinforcement learning approach,” IEEE Commun Mag., vol

sys-55, no 12, Dec 2017

[5] R Huo, F Yu, T Huang, R Xie, J Liu, V Leung, and Y Liu, “Softwaredefined networking, caching, and computing for green wireless networks,”IEEE Commun Mag., vol 54, no 11, pp 185–193, Nov 2016

[6] Q Chen, F R Yu, T Huang, R Xie, J Liu, and Y Liu, “An grated framework for software defined networking, caching and comput-ing,” IEEE Network, vol 31, no 3, pp 46–55, May 2017

inte-[7] C Fang, F R Yu, T Huang, J Liu, and Y Liu, “A survey of greeninformation-centric networking: Research issues and challenges,” IEEEComm Surveys Tutorials, vol 17, no 3, pp 1455–1472, 2015

virtualization over 5G mobile wireless networks,” IEEE Netw., vol 29,

no 3, pp 68–74, May 2015

[9] Z Sanaei, S Abolfazli, A Gani, and R Buyya, “Heterogeneity in mobilecloud computing: Taxonomy and open challenges,” IEEE Commun Surv.Tuts., vol 16, no 1, pp 369–392, 2014

[10] S Andreev, O Galinina, A Pyattaev, J Hosek, P Masek, H.Yanikomeroglu, and Y Koucheryavy, “Exploring synergy between com-munications, caching, and computing in 5G-grade deployments,” IEEECommun Mag., vol 54, no 8, pp 60–69, Aug 2016

[11] M Richart, J Baliosian, J Serrat, and J Gorricho, “Resource slicing

in virtual wireless networks: A survey,” IEEE Trans Netw and ServiceManagement, vol 13, no 3, pp 462–476, Aug 2016

[12] X Zhou, R Li, and T Chen, “Network slicing as a service: enabling terprises own software-defined cellular networks,” IEEE Commun Mag.,vol 54, no 7, pp 146–153, Jul 2016

en-[13] N Bizanis and F Kuipers, “SDN and virtualization solutions for theinternet of things: A survey,” IEEE Access, vol 4, pp 5591–5606, Sept.2016

[14] C Liang and F Yu, “Wireless network virtualization: A survey, someresearch issues and challenges,” IEEE Commun Surv Tuts., vol 17, no

1, pp 358–380, Aug 2014

[15] C Lai, R Hwang, H Chao, M Hassan, and A Alamri, “A buffer-aware

http live streaming approach for sdn-enabled 5g wireless networks,” IEEE

[16] P Agyapong, M Iwamura, D Staehle, W Kiess, and A Benjebbour,

“Design considerations for a 5g network architecture,” IEEE Commun.Mag., vol 52, no 11, pp 65–75, Nov 2014

[17] Q Yan, F R Yu, Q Gong, and J Li, “Software-defined networking(SDN) and distributed denial of service (DDoS) attacks in cloud com-puting environments: A survey, some research issues, and challenges,”IEEE Commun Survey and Tutorials, vol 18, no 1, pp 602–622, 2016.[18] D Kreutz, F Ramos, P Verissimo, C Rothenberg, S Azodolmolky, and

S Uhlig, “Software-defined networking: A comprehensive survey,” ceedings of the IEEE, vol 103, no 1, pp 14–76, Jan 2015

Pro-[19] N McKeown, T Anderson, H Balakrishnan, G Parulkar, L Peterson,

J Rexford, S Shenker, and J Turner, “Openflow: Enabling innovation

in campus networks,” SIGCOMM Comput Commun Rev., vol 38, no

2, pp 69–74, Mar 2008

[20] J Liu, S Zhang, N Kato, H Ujikawa, and K Suzuki, “Device-to-devicecommunications for enhancing quality of experience in software definedmulti-tier lte-a networks,” IEEE Network, vol 29, no 4, pp 46–52, Jul.2015.

[21] L Cui, F R Yu, and Q Yan, “When big data meets software-definednetworking: SDN for big data and big data for SDN,” IEEE Network,vol 30, no 1, pp 58–65, 2016

[22] C Liang and F Yu, “Virtual resource allocation in information-centricwireless virtual networks,” in Proc IEEE International Conference onCommunications (ICC), Jun 2015, pp 3915–3920

[23] J Hoadley and P Maveddat, “Enabling small cell deployment with Net,” IEEE Wireless Commun Mag., vol 19, no 2, pp 4–5, Apr 2012.[24] R Xie, F R Yu, H Ji, and Y Li, “Energy-efficient resource allocationfor heterogeneous cognitive radio networks with femtocells,” IEEE Trans.Wireless Commun., vol 11, no 11, pp 3910–3920, Nov 2012

Het-[25] M Agiwal, A Roy, and N Saxena, “Next generation 5G wireless works: A comprehensive survey,” IEEE Commun Surv Tuts., vol 18,

[28] F Yu and V Krishnamurthy, “Optimal joint session admission control inintegrated WLAN and CDMA cellular networks with vertical handoff,”IEEE Trans Mobile Computing, vol 6, no 1, pp 126–139, Jan 2007.[29] S Bu and F R Yu, “Green cognitive mobile networks with small cellsfor multimedia communications in the smart grid environment,” IEEETrans Veh Tech., vol 63, no 5, pp 2115–2126, Jun 2014

[30] W Cheng, X Zhang, and H Zhang, “Statistical-QoS driven energy ficiency optimization over green 5g mobile wireless networks,” IEEE J.Sel Areas Commun., vol PP, no 99, pp 1–16, Mar 2016, IEEE EarlyAccess Articles

ef-[31] G Huang and J Li, “Interference mitigation for femtocell networks viaadaptive frequency reuse,” IEEE Trans Veh Tech., vol 65, no 4, pp.2413–2423, Apr 2016

IEEE Commun Surv Tuts., vol 17, no 1, pp 405–426, 2015.

[33] Y Cai, F R Yu, and S Bu, “Cloud computing meets mobile wirelesscommunications in next generation cellular networks,” IEEE Network,vol 28, no 6, pp 54–59, Nov 2014

[34] Z Yin, F R Yu, S Bu, and Z Han, “Joint cloud and wireless networksoperations in mobile cloud computing environments with telecom opera-tor cloud,” IEEE Trans Wireless Commun., vol 14, no 7, pp 4020–4033,July 2015

[35] G Fodor, N Rajatheva, W Zirwas et al., “An overview of massive mimotechnology components in metis,” IEEE Commun Mag., vol 55, no 6,

pp 155–161, Jul 2017

[36] K Prasad, E Hossain, and V Bhargava, “Energy efficiency in massivemimo-based 5g networks: Opportunities and challenges,” IEEE WirelessCommun., vol 24, no 3, pp 86–94, Jan 2017

[37] L You, X Gao, G Li, X Xia, and N Ma, “BDMA for wave/terahertz massive mimo transmission with per-beam synchroniza-tion,” IEEE J Sel Areas Commun., vol 35, no 7, pp 1550–1563, Jul.2017

millimeter-[38] L Kong, M Khan, F Wu, G Chen, and P Zeng, “Millimeter-wavewireless communications for iot-cloud supported autonomous vehicles:Overview, design, and challenges,” IEEE Commun Mag., vol 55, no 1,

pp 62–68, Jan 2017

[39] S Han, C I., Z Xu, and C Rowell, “Large-scale antenna systems withhybrid analog and digital beamforming for millimeter wave 5g,” IEEECommun Mag., vol 53, no 1, pp 186–194, Jan 2015

[40] N Ishikawa, R Rajashekar, S Sugiura, and L Hanzo, “Generalizedspatial-modulation-based reduced-rf-chain millimeter-wave communica-tions,” IEEE Trans Veh Technol., vol 66, no 1, pp 879–883, Jan 2017.[41] R Ma, N Xia, H Chen, C Chiu, and C Yang, “Mode selection, radioresource allocation, and power coordination in D2D communications,”IEEE Wireless Commun., vol 24, no 3, pp 112–121, Feb 2017

[42] Y Cai, F R Yu, C Liang, B Sun, and Q Yan, “Software defined to-device (D2D) communications in virtual wireless networks with imper-fect network state information (NSI),” IEEE Trans Veh Tech., no 9, pp.7349–7360, Sept 2016

device-[43] K Wang, F Yu, H Li, and Z Li, “Information-centric wireless networkswith virtualization and D2D communications,” IEEE Wireless Commun.,vol 24, no 3, pp 104–111, Jan 2017

[44] Y Zhou and W Yu, “Optimized backhaul compression for uplink cloudradio access network,” IEEE J Sel Areas Commun., vol 32, no 6, pp.1295–1307, Jun 2014.

[45] A Ioannou and S Weber, “A survey of caching policies and forwardingmechanisms in information-centric networking,” IEEE Commun Surv.Tuts., vol PP, no 99, pp 1–41, 2016, IEEE Early Access Articles.[46] ETSI, “Mobile-edge computing: Introductory technical white paper,”ETSI White Paper, Sept 2014

[47] V Jacobson, D Smetters, J Thornton, M Plass, N Briggs, and R nard, “Networking named content,” in Proc 5th International Conference

Bray-on Emerging Networking Experiments and Technologies (CoNEXT09),Rome, Italy, Dec 2009, pp 1–12

[48] R Buyya and M Pathan, Content Delivery Network Berlin: Springer,2008

[49] S Androutsellis and D Spinellis, “A survey of peer-to-peer content tribution technologies,” ACM Computing Surveys (CSUR), vol 36, no

dis-4, pp 335–371, Dec 2004

[50] A Passarella, “A survey on content-centric technologies for the currentinternet: CDN and p2p solutions,” Elsevier Journal on Computer Com-munications, vol 35, no 1, pp 1–32, Jan 2011

[51] E Lua, J Crowcroft, M Pias, R Sharma, and S Lim, “A survey andcomparison of peer-to-peer overlay network schemes,” IEEE Commun.Surv Tuts., vol 7, no 2, pp 72–93, Apr 2005

[52] B Ahlgren, C Dannewitz, C Imbrenda, D Kutscher, and B Ohlman,

“A survey of information-centric networking,” IEEE Commun Mag., vol

50, no 7, pp 26–36, July 2012

[53] G Xylomenos, C Ververidis, V Siris, N Fotiou, C Tsilopoulos, X lakos, K Katsaros, and G Polyzos, “A survey of informationcentric net-working research,” IEEE Commun Surv Tuts., vol 16, no 2, pp 1–26,July 2013

Vasi-[54] K Djemame, D Armstrong, J Guitart, and M Macias, “A risk ment framework for cloud computing,” IEEE Trans Cloud Computing,vol 4, no 3, pp 265–278, Jul 2016

assess-[55] A Khan, M Othman, S Madani, and S Khan, “A survey of mobilecloud computing application models,” IEEE Commun Surv Tuts., vol

16, no 1, pp 393–413, 2014

[56] M Zhanikeev, “A cloud visitation platform to facilitate cloud federationand fog computing,” Computer, vol 48, no 5, pp 80–83, May 2015